Correlated, Yet Independent A Correlation Based Feature Selection Algorithm

Feature selection is a critical process in machine learning and data analysis, aiming to identify the most relevant subset of features from a dataset and thereby improve model performance, interpretability, and computational efficiency. However, with the recent increase in data dimensionality, the complexity and dependency on model performance have become common among many existing methods, which has become a disadvantage. To address these issues, the current research paper introduces an improved correlation-based feature selection method that optimizes non-redundant relevant feature subset selection. The study evaluates this method alongside other established feature selection methods using real-world and benchmark datasets across various experiments. The findings show that the feature subset generated b y the proposed algorithm corresponds to those identified by alternative methods. The algorithm also consistently identifies a compact feature set, yielding notable predictive performance on unseen test data, and statistically significant execution time. These results highlight the effectiveness and potential of enhancing feature selection processes for diverse applications.